New RJ interview on self-driving: using LiDAR to catch up to Tesla (+ much more topics)

[Katsudon]

"Tesla's s got a huge fleet. So our approach is we have to catch up with them and compete. And so we're going to have more cameras with better dynamic range. We're going to supplement that with LiDAR that provides better safety for edge case conditions and allows us to train the vision models faster. And the incremental cost to do that is relatively small.

And in fact any of the incremental costs that would have been there on its own is offset by the fact that we brought inference in house and reduced the cost of our inference platform so dramatically from what we have in your car. Your car uses an Nvidia inference platform. I say all this because in like the infinite long-term you know, you could make the case that once the models are very very robust you could have less cameras or you may be able to get away with less radar. It's not clear yet if that's the case for covering all these corner cases."



Chapters:
0:00 - Intro
2:45 - Rivian’s big 2026 moment
4:23 - The origin of Rivian
6:25 - The pivot that changed Rivian
7:21 - Rivian’s core mission
9:06 - Obsessing over details
9:57 - Why R2 matters
12:07 - Cutting cost, keeping quality
15:21 - One brain, thousands of decisions
18:56 - Rivian’s software advantage
19:02 - Autonomy and the physical world
20:28 - The AI shift in self-driving
23:26 - Rivian’s autonomy roadmap
25:26 - Training AI from real driving
28:41 - R2 as a data machine
29:45 - Vision vs LiDAR
35:43 - Safety and corner cases
37:35 - Fewer cars or more driving?
40:03 - Robotaxis vs car ownership
42:21 - RJ’s robotics thesis beyond the humanoid hype
47:59 - How to raise kids for an unrecognizable future
50:40 - The timeline that should worry everyone

Sponsored

 

[VandalSibs]

I'm skimming thru a lot of this, and it just makes me appreciate RJ's though process even more. He seems to actually think long-term and drive the company that way, as opposed to others that constantly try to pivot at every single new thing.

His comments towards the end about emphasizing the need for more empathy in the world and in government in particular were quite heartening to hear.

 

[Billyk24]

Lidar or other lack of it on the Launch edition is a bummer. I am waiting for Lidar to arrive with its price tag.

 

[TexasBob]

The Tesla "huge fleet" nonsense rears its head. Tesla has a large-ish (5 million) fleet of camera only data feeds. (A lot of the older vehicles are not capable of useful data gathering because computing is too limited or the resolution is not good). It has a tiny fleet that combines high definition radar and camera data (2024+ model S/X). This sensor fusion will, IMO, be required and because the rest of Tesla is camera only, the data set is weak. (It is good compared to Rivian but not exactly a relevant benchmark.)

Compare that to the 170 million Mobileye vehicles on the road or the 8 million that are capable and actively being used for multi-sensor data harvesting. That is why they can do a complete country scale HD map in weeks.

The real game is high quality simulation built on extremely high quality multi-sensor real-world data and on that front NVIDIA is also providing that full capability to its customers (Lucid/Uber/Nuro, etc.). Rivian can absolutely do this, but it will be slower and less effective than buying it from Mobileye or NVIDIA of qualcomm. It will be cheaper in the long run and you can control the timeline more but sucks resources in the near term.

Sometimes I wonder if (a) RJ really believes the Tesla story in which case smh or (b) is just avoiding Elon's market-moving hate-tweets in which case I get it of (c) is trying to convince the retail investor rubes to throw foolish meme-stock money at Rivian in which case, shame on him.

 

[MountainBikeDude]

Lidar or other lack of it on the Launch edition is a bummer. I am waiting for Lidar to arrive with its price tag.

More and more I feel like LiDAR inclusion is more for Rivian's benefit in training models, rather than something a user will notice from a driving or future proofing standpoint between R2 Gen 2 Autonomy platform without, and R2 Gen 3 Autonomy platform with.

 

[UnsungZero_OldTimeAdMan]

More and more I feel like LiDAR inclusion is more for Rivian's benefit in training models, rather than something a user will notice from a driving or future proofing standpoint between R2 Gen 2 Autonomy platform without, and R2 Gen 3 Autonomy platform with.

Exactly what it is. And the often repeated reasons to wait for it, IMO, are misinformed and fueled by FOMO. Level 4 autonomy requires legislation and govt. approval. The path for that is full of legal debate. It won't happen anytime soon, if at all. Until there is legal approval, LiDAR (and those who have it) will just be part of a development process; i.e. unpaid developer/contributor... mapping roads and providing data for training AI. Get the car you want, or need, now. FOMO is irrational, since this is likely not the last car you will be buying for the rest of your life. Resale value worries? Pffft. Every single car, LiDAR or not, is a depreciating asset. And Teslas with or without FSD has already proven it doesn't factor into resale significantly. Vehicle condition, warranty and battery health are greater factors (i.e. all the usual resale things).

 

[captainjp]

More and more I feel like LiDAR inclusion is more for Rivian's benefit in training models, rather than something a user will notice from a driving or future proofing standpoint between R2 Gen 2 Autonomy platform without, and R2 Gen 3 Autonomy platform with.

I was thinking the same. I don’t believe the system will rely on it as much as one thinks after the bulk of ground truth data is compiled.

 

[JoulesVerne]

RJ says the benefit is in edge cases, which to me, translates to avoiding collisions (or disengagement events) in cases where vision-only fails. That seems like a clear benefit to the driver of a LIDAR equipped vehicle.

 

[DuoRivian]

I was thinking the same. I don’t believe the system will rely on it as much as one thinks after the bulk of ground truth data is compiled.

I am also wondering why people talk of level 2 and level 4 autonomy and skip right over level 3. It does seem like the benefits of lidar in the next few years is limited for any customer and would be useful in a few instances.

 

[skyguyscott]

RJ says the benefit is in edge cases, which to me, translates to avoiding collisions (or disengagement events) in cases where vision-only fails. That seems like a clear benefit to the driver of a LIDAR equipped vehicle.

^^^^THIS^^^^

 

[NeverFollow]

The Lidar situation looks a little bit similar to Volvo who decided to drop LIDAR as a standard feature for its 2026 model year electric vehicles (primarily the EX90 and ES90), to meet customer demand and improve vehicle availability, and avoid LIDAR delays and production issues .

Volvo announced that its new, powerful "core computing" systems, combined with a robust suite of cameras and radar, can deliver high-level safety and driver assistance without needing LIDAR.

So Rivian is playing the safe card approach by providing a highly capable camera-and-radar-based ADAS, lacking LiDAR initially, but featuring an "elevated" Gen 2 compute platform sufficient for most driver needs.

 

[Donald Stanfield]

I'm skimming thru a lot of this, and it just makes me appreciate RJ's though process even more. He seems to actually think long-term and drive the company that way, as opposed to others that constantly try to pivot at every single new thing.

His comments towards the end about emphasizing the need for more empathy in the world and in government in particular were quite heartening to hear.

I agree. RJ is making moves to position Rivian as a long-term player in the auto manufacturing space.

Sponsored